Hadi Valipour
AI and Data Science Leader
Executive Profile
Hands-on AI and Data Science leader operating at the frontier of applied AI, designing and building production-grade agentic systems, LLM architectures, and evaluation frameworks that are safe, reliable, and scalable. Currently a Senior Manager at Mable (Attain Healthtech), where I retain direct architectural ownership of AI systems: from agent orchestration patterns and context management strategies through to evaluation methodology, modelling decisions, and algorithmic design.
Core Strengths
- Agentic AI architecture and LLM systems: orchestration, tool use, context management, RAG, and reasoning architecture.
- AI evaluation methodology: automated and human evaluation harnesses, online experimentation, offline metrics, and safety guardrails.
- Production AI and MLOps: CI/CD, canarying, shadow traffic, drift monitoring, observability, and incident playbooks.
- AI strategy and roadmap ownership with executive stakeholders across investment, prioritisation, and delivery alignment.
- Team leadership and org design for multidisciplinary AI and Data Science teams.
- Responsible AI and safety governance: risk classification, human-in-the-loop systems, model reviews, and failure-mode analysis.
- Platform data science and experimentation across matching, ranking, personalisation, churn, and representation learning.
Working Experience
Senior Manager, Data Science & AI - Mable
Jan 2026 - Present
Melbourne, Australia
- Architect agentic AI systems end to end: orchestration patterns, reasoning architecture, tool-use strategies, context management, and prompt engineering.
- Own evaluation methodology for LLM and agentic systems, including automated and human evaluation harnesses, helpfulness metrics, safety metrics, and A/B experimentation.
- Lead a multidisciplinary AI and data team across Attain Healthtech brands while retaining hands-on ownership of architecture, modelling, algorithms, and evaluation.
- Own AI governance, risk classification, evaluation standards, and compliance processes across the organisation.
- Partner with executive leadership on AI investment decisions, portfolio prioritisation, and long-term platform strategy.
Manager, Data Science & AI - Mable
May 2022 - Dec 2025
Melbourne, Australia
- Designed and delivered production AI systems across matching/ranking, personalisation, retention, agentic customer assistance, and workflow automation.
- Designed and implemented a GenAI agentic system for customer assistance, including orchestration architecture, RAG pipeline, evaluation framework, and safety guardrails.
- Developed a contrastive representation learning model for job-provider matching and a deep multivariate time-series model for churn prediction.
- Led MLOps and platform capability across CI/CD, drift monitoring, observability, incident playbooks, and model governance reviews.
- Optimised latency, throughput, and cost across real-time AI systems on Databricks and GCP.
Director of Engineering and Data Science - Ostadkar
Jul 2019 - Mar 2022
Tehran, Iran
- Built and led Engineering, Data Science, and Analytics organisation (30+ team, 8 direct reports).
- Owned end-to-end AI/ML system architecture: adaptive pricing, recommendation systems, churn and fulfilment prediction, and intelligent ticket allocation.
- Designed experimentation frameworks for dynamic pricing, revenue optimisation, and retention systems.
- Led monolith-to-microservices transformation and improved system reliability, reducing MTTR by about 50%.
- Impact: take-rate +6 to +10 pp, fulfilment +20%, provider churn down 30%, lead time down 25%.
Engineering and Data Science Lead - SIB Team
Dec 2013 - Jul 2019
Tehran, Iran
- Led engineering and data teams (15+) and delivered a service-delivery platform supporting about 1M daily transactions.
- Built ML systems for collaborative filtering, retention, recommendation, forecasting, and sentiment analysis.
- Impact: bundle-suggest success up about 40%; telecom retention improved by +20 pp via collaborative filtering.
Senior Software Engineer / Data Scientist - ParsLib
Feb 2012 - Dec 2013
Tehran, Iran
- Built service-delivery and content-pool platforms with production NLP and CV applications.
- Scaled content systems to ~100K items and supported both development and infrastructure operations.
Earlier Roles and Research Experience
2008 - 2012
- Software Engineer (Part-time), ParsLib
- Researcher, Multimedia Processing Lab
- Visiting Researcher, Cognitive Robotics Lab
- Software Development Intern, IranAir
Education
Master of Computer Science - Artificial Intelligence
Amirkabir University of Technology (Tehran Polytechnic), Tehran, Iran (2009 - 2012)
Technical Toolbox
- LLMs & Agentic AI: LangGraph, tool use, multi-agent orchestration, prompt and context engineering, RAG, fine-tuning, evaluation, safety guardrails.
- ML & Modelling: representation learning, deep learning, learning-to-rank, time-series analysis, collaborative filtering, probabilistic and Bayesian methods.
- Experimentation & Evaluation: A/B/n, CUPED, online/offline evaluation, latency, fairness and abuse guardrails, telemetry-driven iteration.
- MLOps & Reliability: CI/CD, canarying, shadow traffic, drift monitoring, SLOs, incident playbooks, observability, model lifecycle management.
- Architecture & Platforms: microservices and APIs, MCP, feature stores, data contracts, event-driven systems.
- Cloud & Tooling: Python, SQL, PySpark/Spark, Databricks, MLflow, AWS/GCP.
Selected AI Projects & Initiatives
Agentic AI System for Workflow Automation (Attain Healthtech)
2025Designed and built a GenAI agentic system integrated with internal subsystems to triage emails, generate responses, and execute actions autonomously and safely.
GenAI Agentic Assistant for Customer Guidance (Mable)
2024 - 2025Architected and delivered a personalised conversational assistant, including orchestration design, RAG pipeline, evaluation harness, and safety guardrails.
AI-Based Shift Notes Moderation (Attain Healthtech)
2023 - 2024Fine-tuned and evaluated a custom LLM with clinical expertise to review shift notes and flag potential welfare issues at scale.
Client Churn Reduction with Proactive Intervention (Mable)
2024Built a deep multivariate time-series classification model for churn prediction and a proactive intervention pipeline for at-risk clients.
Job-Provider Matching Optimization (Mable)
2022 - 2023Built a contrastive representation learning model to predict successful job-provider matches and improve marketplace efficiency.
Adaptive Pricing and Revenue Forecasting (Ostadkar)
2021Developed an adaptive pricing model using predictive analytics to dynamically select revenue models and adjust pricing.
Intelligent Ticketing System with Predictive Allocation (Ostadkar)
2021Built intelligent ticket generation powered by predictive algorithms, improving fulfilment by 20%.
Selected Publications
- Enhanced Job Exposure with Contrastive Person-Job Fit Matching in Job Marketplaces (in progress).
- Customer Churn Prediction Using Contrastive Learning and Representation Trajectory (in progress).
- GHRS: Graph-based Hybrid Recommendation System, Expert Systems with Applications, 2022.
- Optimization of Emotional Learning Approach to Control Systems with Unstable Equilibrium, LNAI, 2015.
- A Brief Survey of Software Architecture Concepts and Service-Oriented Architecture, IEEE ICCSIT, 2009.
Awards & Memberships
- Outstanding Paper, IEEE/ACIS SNPD 2014; Best Paper, CSI 2012.
- Honourable Mention (twice), ICPC ACM Asia Regional Programming Contest, 2005 and 2006.
- Reviewer, Engineering Applications of AI (Elsevier), Scientific Reports (Nature), IEEE JBHI, and multiple conferences.
CV
Download the latest CV: Hadi Valipour - CV.